Spatiotemporal Knowledge-Driven Evolutionary Algorithm for Dynamic Constrained Multiobjective Optimization

Dynamic constrained multiobjective optimization problems (DCMOPs) present significant challenges due to timevarying objectives and constraints. To address these challenges, we propose a spatiotemporal knowledge-driven evolutionary algorithm (SKEA) that leverages both historical and current environme...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2025 7th International Conference on Data-driven Optimization of Complex Systems (DOCS) S. 271 - 276
Hauptverfasser: Liu, Zhi, Song, Wei
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 19.08.2025
Schlagworte:
Current measurement
differential evolution
dynamic constrained multiobjective optimization
environmental match
Evolutionary computation
Heuristic algorithms
Knowledge transfer
Neural networks
Optimization
Prediction algorithms
Robustness
Spatiotemporal phenomena
Transfer learning
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Dynamic constrained multiobjective optimization problems (DCMOPs) present significant challenges due to timevarying objectives and constraints. To address these challenges, we propose a spatiotemporal knowledge-driven evolutionary algorithm (SKEA) that leverages both historical and current environmental knowledge to guide the search process. Specifically, temporal knowledge is acquired from the most similar historical environment by measuring changes in objective values and constraint violations. Based on this knowledge, the main and auxiliary populations are reconstructed to adapt quickly to the new environment. Furthermore, the populations are divided into four categories according to feasibility and dominance, including nondominated feasible solutions, dominated feasible solutions, nondominated infeasible solutions, and dominated infeasible solutions. By constructing valuable solution pairs among these categories, four single hidden layer neural networks are trained to predict high-quality learning targets for each type of solution. These predicted targets are then used to guide the mutation operation in a modified differential evolution algorithm, while category-based crossover enhances local exploration. The experimental results on various benchmarks verify that, compared with several state-of-the-art algorithms, the proposed SKEA achieves the most competitive performance in solving DCMOPs.
DOI:10.1109/DOCS67533.2025.11200903